REACT-1 round 10 report: Level prevalence of SARS-CoV-2 swab-positivity in Englandduring third national lockdown in March 2021

Steven Riley1,2,*, Oliver Eales1,2, David Haw1,2, Caroline E. Walters1,2, Haowei Wang1,2, Kylie

E. C. Ainslie1,2,12, Christina Atchison1, Claudio Fronterre3, Peter J. Diggle3, Deborah Ashby1,

Christl A. Donnelly1,2,4, Graham Cooke5,6,7, Wendy Barclay5, Helen Ward1,6,7, Ara Darzi6,7,8,

Paul Elliott1,6,7,9,10,11 *

1 School of Public Health, Imperial College London, UK

2 MRC Centre for Global infectious Disease Analysis and Abdul Latif Jameel Institute for

Disease and Emergency Analytics, Imperial College London, UK

3 CHICAS, Lancaster Medical School, Lancaster University, UK and Health Data Research,

UK

4 Department of Statistics, University of Oxford, UK

5 Department of Infectious Disease, Imperial College London, UK

6 Imperial College Healthcare NHS Trust, UK

7 National Institute for Health Research Imperial Biomedical Research Centre, UK

8 Institute of Global Health Innovation at Imperial College London, UK

9 MRC Centre for Environment and Health, School of Public Health, Imperial College

London, UK

10 Health Data Research (HDR) UK London at Imperial College

11 UK Dementia Research Institute at Imperial College

12 Centre for Infectious Disease Control, National Institute for Public Health and the

Environment, Bilthoven, The Netherlands

*Corresponding authors: Steven Riley and Paul Elliott, s.riley@imperial.ac.uk,p.elliott@imperial.ac.uk, School of Public Health, Imperial College London, Norfolk Place,London, W2 1PG

1

Abstract

Background

In England, hospitalisations and deaths due to SARS-CoV-2 have been falling consistentlysince January 2021 during the third national lockdown of the COVID-19 pandemic. The firstsignificant relaxation of that lockdown occurred on 8 March when schools reopened.

Methods

The REal-time Assessment of Community Transmission-1 (REACT-1) study augmentsroutine surveillance data for England by measuring swab-positivity for SARS-CoV-2 in thecommunity. The current round, round 10, collected swabs from 11 to 30 March 2021 and iscompared here to round 9, in which swabs were collected from 4 to 23 February 2021.

Results

During round 10, we estimated an R number of 1.00 (95% confidence interval 0.81, 1.21).Between rounds 9 and 10 we estimated national prevalence has dropped by ~60% from0.49% (0.44%, 0.55%) in February to 0.20% (0.17%, 0.23%) in March. There weresubstantial falls in weighted regional prevalence: in South East from 0.36% (0.29%, 0.44%)in round 9 to 0.07% (0.04%, 0.12%) in round 10; London from 0.60% (0.48%, 0.76%) to0.16% (0.10%, 0.26%); East of England from 0.47% (0.36%, 0.60%) to 0.15% (0.10%,0.24%); East Midlands from 0.59% (0.45%, 0.77%) to 0.19% (0.13%, 0.28%); and NorthWest from 0.69% (0.54%, 0.88%) to 0.31% (0.21%, 0.45%). Areas of apparent higherprevalence remain in parts of the North West, and Yorkshire and The Humber. The highestprevalence in March was found among school-aged children 5 to 12 years at 0.41% (0.27%,0.62%), compared with the lowest in those aged 65 to 74 and 75 and over at 0.09% (0.05%,0.16%). The close approximation between prevalence of infections and deaths (suitablylagged) is diverging, suggesting that infections may have resulted in fewer hospitalisationsand deaths since the start of widespread vaccination.

Conclusion

We report a sharp decline in prevalence of infections between February and March 2021.We did not observe an increase in the prevalence of SARS-CoV-2 following the reopening ofschools in England, although the decline of prevalence appears to have stopped. Futurerounds of REACT-1 will be able to measure the rate of growth or decline from this currentplateau and hence help assess the effectiveness of the vaccination roll-out on transmissionof the virus as well as the potential size of any third wave during the ensuing months.

2

Introduction

During the spring of 2021, many European populations are suffering from substantial third

waves of the COVID-19 pandemic with high pressure on health care systems [1] resulting in

stringent social distancing. In England, during the third national lockdown, hospitalisations

and deaths have fallen consistently since early January [2]. However, since the first

substantial relaxation of lockdown in England with the opening of schools on 8 March 2021

[3], the rate of decline of new cases has slowed considerably [2]. Large-scale testing of

school children and their families using lateral flow devices accompanied the opening of

schools [4].

The REal-time Assessment of Community Transmission-1 (REACT-1) study augments

routine surveillance data for England by measuring swab-positivity for SARS-CoV-2 in a

random sample of the community recruited regardless of symptom status [5]. We present

here the results from round 10 of REACT-1 from swabs collected from 11 to 30 March 2021.

We compare the results of REACT-1 round 10 to round 9, in which swabs were collected

from 4 to 23 February 2021.

Results

In round 10 we found 227 positives from 140,844 swabs giving an unweighted prevalence of

0.16% (0.14%, 0.18%) and a weighted prevalence of 0.20% (0.17%, 0.23%) (Table 1). This

represents a ~60% reduction from round 9 in which unweighted prevalence was 0.42%

(0.39%, 0.45%) and weighted prevalence was 0.49% (0.44%, 0.55%).

Using a constant growth rate model, on average for England, we found evidence for a

decline over the period of round 9 to 10 (Table 2, Figure 1) with an estimated halving time of

26 (23, 31) days and a corresponding R of 0.84 (0.82, 0.86). Within round 10, we estimate

an R of 1.00 (0.81, 1.21) with 51% probability that R > 1. These results suggest a slow down

in the rate of decline, with prevalence plateauing at around 1 in 500 during round 10 (Figure

1). These results were largely unchanged in sensitivity analyses in which subsets of positive

samples were considered: non-symptomatics, only participants with swabs positive for both

E and N genes, and only those positive for N gene with a CT value of 35 or below. However,

we estimated a lower R of 0.72 (0.50, 0.99) for round 10 based only on the prevalence of

positive swabs taken from non-symptomatic participants (Table 2).

At regional level, R over the period of rounds 9 to 10 ranged from 0.75 (0.68, 0.81) in South

East to 0.95 (0.87, 1.03) in Yorkshire and The Humber. We estimated a 12% probability that

3

R > 1 between rounds 9 and 10 in Yorkshire and the Humber, while the probability of R >1

was less than 1% in all other regions (Table 3).

Reflecting these regional R numbers, there were substantial falls in weighted regional

prevalence: in South East from 0.36% (0.29%, 0.44%) in round 9 to 0.07% (0.04%, 0.12%)

in round 10; London from 0.60% (0.48%, 0.76%) to 0.16% (0.10%, 0.26%); East of England

from 0.47% (0.36%, 0.60%) to 0.15% (0.10%, 0.24%); East Midlands from 0.59% (0.45%,

0.77%) to 0.19% (0.13%, 0.28%); and North West from 0.69% (0.54%, 0.88%) to 0.31%

(0.21%, 0.45%) (Table 4, Figure 2).

At sub-regional scale, areas of apparent higher prevalence remain in round 10 in parts of

North West, and Yorkshire and The Humber (Figure 3).

We found the highest weighted prevalence of swab-positivity in those aged 5 to 12 years at

0.41% (0.27%, 0.62%) compared with the lowest in those aged 65 to 74 years and 75 years

and over at 0.09% (0.05%, 0.16%) (Table 4, Figure 4). As with prior rounds, we found much

higher prevalence of swab-positivity during round 10 in those who had come into contact

with a confirmed COVID-19 case at 4.9% (3.4%, 7.0%) compared with those who had not at

0.14% (0.11%, 0.17%) (Table 4a). Also, those living in the least deprived neighbourhoods

had a lower prevalence of swab-positivity at 0.13% (0.09%, 0.17%) compared with 0.33%

(0.23%, 0.47%) for those in the most deprived neighbourhoods (Table 4a).

Frequency of foreign travel remained low overall during round 10 compared with prior

rounds of REACT-1 (Figure 5). Foreign travel remains higher in London compared with other

regions except for Yorkshire and The Humber.

We investigated the relationship between swab-positivity, as estimated in rounds 1 to 8 of

REACT-1, and hospital admissions (Figure 6), and deaths (Figure 7) from publicly available

national data. We estimated lags of 18 (18, 18) days between the timing of swab results and

hospital admissions, and 27 (27, 27) days between swab and death to give the best fit to

REACT-1 rounds 1 to 8 prevalence data. During the period of round 9, curves for deaths and

(more marginally) for hospital admissions diverged from the prevalence P-spline, suggesting

that infections may have resulted in fewer hospitalisations and deaths since the start of

widespread vaccination.

Discussion

In this tenth round of the REACT-1 study carried out from 11 to 30 March 2021, we report an

approximate 60% reduction in prevalence since the previous round from 4 to 23 February.

These rounds occurred during the third national lockdown in England [6] following the

4

second wave of infections, which grew rapidly through December 2020 until the beginning of

January 2021. The most recent round also covered the start of the gradual easing of

lockdown [3], notably the opening up of schools from 8 March 2021.

Our data indicate a slowing of the epidemic from February to March, with the national R

number over that period reliably below one. However, in the most recent data for March, the

rate of decline slowed and prevalence flattened at 1 in 500 (0.20%) people infected overall,

with prevalence highest in primary and early secondary school-aged children (5 to 12 years)

and lowest in those above the age of 65 years. The lower rate among older adults is

consistent with an effect from the vaccination roll-out in England, that has focused initially on

older people and the most vulnerable.

Our analysis of mortality and to a lesser extent hospitalisation trends is consistent with the

vaccination programme contributing in part to lower rates of severe outcomes in the older

age group. In recent rounds of REACT-1 there was a close approximation between

prevalence of infection (measured by REACT-1) and hospitalisation and mortality rates (from

publicly available national data sources, after considering a suitable lag period between

infection and outcome). This relationship has diverged, particularly for deaths, from late

January onwards and is in keeping with the timing of the roll-out of the mass vaccination

programme. This observation is consistent with studies in Scotland [7] and Israel [8] which

reported high efficacy of vaccination against severe disease. It will be important to monitor

future trends in the relationship between infections, deaths and hospitalisations in England.

Specifically if the approximation of the infection and mortality curves revert to prior patterns,

this might indicate an increase in frequency of variants associated with reduced efficacy of

the vaccine. We note that during early rounds of REACT-1 the relationship between

infections and mortality appeared weaker: this likely reflects a higher proportion of infections

occurring in health and care facilities at that time rather than in the community later in the

epidemic [9].

Swab-positivity was over 30-fold higher in participants who reported contact with a confirmed

case compared with those who did not. This reinforces the high potential value of testing

known contacts even though manual contact tracing alone may not be sufficient to achieve

R < 1 in the absence of other interventions [10,11]. Further, the consistently higher rates of

infection in the most deprived neighbourhoods observed in REACT-1 suggest a potential

role for targeted highly-local interventions during periods of lower prevalence, even when the

presence of a novel variant is not a primary concern [12].

Prevalence of infection in England is now at a level last seen in early September 2020, and

nearing the prevalence at the end of the first lockdown in England in May 2020. The

5

Government in England has adopted a cautionary approach to easing of lockdown with

approximately five week intervals between each step toward ‘normalisation’ of activities

including opening of non-essential shops, indoor sports, outdoor hospitality and personal

services (such as hairdressing) from 12 April 2021 [3]. This approach allows time to assess

the effect of each step on the transmission of the virus and the R number, which will be

measured in subsequent REACT-1 surveys, among others, over the ensuing period.

Meanwhile, our estimate of 1.0 for the R number in March suggests that the fine balance

between the cautious opening up of society and the need to control the levels of infection

has not so far resulted in a third wave of infections. The continued roll-out of the vaccination

programme, including second doses, and the extension to adult ages below 50 years will

further increase immunity in the population.

Methods

REACT-1 involves a random sample of the population in England (ages 5 years and over)

who are invited to undertake a self-administered throat and nose swab in the home

(parent/guardian administered at ages 5 to 12 years) and complete a health and lifestyle

questionnaire [5]. We request that completed swabs are refrigerated before being

transported by courier and sent chilled to the laboratory for RT-PCR testing.

We obtain the sample using the National Health Service patient register and aim to achieve

approximately equal numbers of participants in each of the 315 lower tier local authorities

(LTLAs) in England. In round 10, we sent out 760,000 letters to named individuals and

dispatched 180,106 (23.7%) test kits, resulting in 140,844 (78.2%) completed swabs with a

valid test result. This gave an overall response rate (valid swabs divided by total number of

letters sent) of 18.5%.

We obtain prevalence of RT-PCR swab-positivity with and without weighting. This is done

both to take account of the sample design (balanced by LTLA, not by population) and to

adjust for variable non-response. The aim is to provide prevalence estimates (national,

regional, and by specific demographic and occupational groups) that are representative of

the population of England as a whole, by age, sex, region and ethnicity. We also obtain a

map of smoothed prevalence at LTLA level by use of a neighbourhood spatial smoothing

method across nearest neighbours up to 30 km.

We use exponential growth models to estimate the reproduction number R, across and

within rounds. Across all rounds, we fit a smoothed P-spline function to the daily prevalence

data, with knots at 5-day intervals, to examine trends in unweighted prevalence over time

[13]. We compare the daily prevalence data in REACT-1 with the publicly available national

6

daily hospital admissions and COVID-19 mortality data (deaths within 28 days of a positive

test), with allowance for a suitable discrete-day lag period between infection and either

hospital admission or death and an appropriate scaling parameter.

In sensitivity analyses for the estimation of R, we provide alternative cut-points of cycle

threshold (CT) values for swab-positivity and restrict the analyses only to those who did not

report symptoms in the previous week.

Statistical analyses were carried out in R [14]. We obtained research ethics approval from

the South Central-Berkshire B Research Ethics Committee (IRAS ID: 283787).

References

1. Coronavirus Pandemic (COVID-19). In: Our World in Data [Internet]. [cited 6 Apr 2021].Available: https://ourworldindata.org/

2. UK Government. UK government Covid-19 dashboard. In: UK government Covid-19dashboard [Internet]. Available: https://coronavirus.data.gov.uk/

3. UK Government. Prime Minister sets out roadmap to cautiously ease lockdownrestrictions. In: GOV.UK [Internet]. [cited 3 Feb 2021]. Available:https://www.gov.uk/government/news/prime-minister-sets-out-roadmap-to-cautiously-ease-lockdown-restrictions

4. UK Government. Schools, nurseries and colleges testing: order coronavirus (COVID-19)rapid lateral flow home test kits. In: uk.gov [Internet]. [cited 6 Apr 2021]. Available:https://www.gov.uk/order-coronavirus-rapid-lateral-flow-tests

5. Riley S, Atchison C, Ashby D, Donnelly CA, Barclay W, Cooke G, et al. REal-timeAssessment of Community Transmission (REACT) of SARS-CoV-2 virus: Studyprotocol. Wellcome Open Research. 2020. p. 200.doi:10.12688/wellcomeopenres.16228.1

6. UK Government. Prime Minister announces national lockdown. In: gov.uk [Internet]. 4Jan 2021 [cited 16 Feb 2021]. Available:https://www.gov.uk/government/news/prime-minister-announces-national-lockdown

7. Torjesen I. Covid-19: First doses of vaccines in Scotland led to a substantial fall inhospital admissions. BMJ. 2021;372: n523.

8. Dagan N, Barda N, Kepten E, Miron O, Perchik S, Katz MA, et al. BNT162b2 mRNACovid-19 Vaccine in a Nationwide Mass Vaccination Setting. N Engl J Med. 2021.doi:10.1056/NEJMoa2101765

9. Iacobucci G. Covid-19: Doctors sound alarm over hospital transmissions. BMJ.2020;369: m2013.

10. Kucharski AJ, Klepac P, Conlan AJK, Kissler SM, Tang ML, Fry H, et al. Effectiveness ofisolation, testing, contact tracing, and physical distancing on reducing transmission ofSARS-CoV-2 in different settings: a mathematical modelling study. Lancet Infect Dis.2020;20: 1151–1160.

7

11. Ferretti L, Wymant C, Kendall M, Zhao L, Nurtay A, Abeler-Dörner L, et al. QuantifyingSARS-CoV-2 transmission suggests epidemic control with digital contact tracing.Science. 2020;368. doi:10.1126/science.abb6936

12. UK Government. Surge testing for new coronavirus (COVID-19) variants. In: gov.uk[Internet]. 2 Oct 2021 [cited 6/ 4/ 2021]. Available:https://www.gov.uk/guidance/surge-testing-for-new-coronavirus-covid-19-variants

13. Lang S, Brezger A. Bayesian P-Splines. J Comput Graph Stat. 2004;13: 183–212.

14. R Core Team. R: A Language and Environment for Statistical Computing. R Foundationfor Statistical Computing; 2020. Available: https://www.R-project.org/

Data availability

Supporting data for tables and figures are available either: in this spreadsheet; or in theinst/extdata directory of this GitHub R package.

Declaration of interests

We declare no competing interests.

Funding

The study was funded by the Department of Health and Social Care in England.

Acknowledgements

SR, CAD acknowledge support: MRC Centre for Global Infectious Disease Analysis,

National Institute for Health Research (NIHR) Health Protection Research Unit (HPRU),

Wellcome Trust (200861/Z/16/Z, 200187/Z/15/Z), and Centres for Disease Control and

Prevention (US, U01CK0005-01-02). GC is supported by an NIHR Professorship. HW

acknowledges support from an NIHR Senior Investigator Award and the Wellcome Trust

(205456/Z/16/Z). PE is Director of the MRC Centre for Environment and Health

(MR/L01341X/1, MR/S019669/1). PE acknowledges support from Health Data Research UK

(HDR UK); the NIHR Imperial Biomedical Research Centre; NIHR HPRUs in Chemical and

Radiation Threats and Hazards, and Environmental Exposures and Health; the British Heart

Foundation Centre for Research Excellence at Imperial College London (RE/18/4/34215);

and the UK Dementia Research Institute at Imperial (MC_PC_17114). We thank The Huo

Family Foundation for their support of our work on COVID-19.

We thank key collaborators on this work – Ipsos MORI: Kelly Beaver, Sam Clemens, Gary

Welch, Nicholas Gilby, Kelly Ward and Kevin Pickering; Institute of Global Health Innovation

at Imperial College: Gianluca Fontana, Sutha Satkunarajah, Didi Thompson and Lenny

Naar; Molecular Diagnostic Unit, Imperial College London: Prof. Graham Taylor; North West

London Pathology and Public Health England for help in calibration of the laboratory

8

analyses; Patient Experience Research Centre at Imperial College and the REACT Public

Advisory Panel; NHS Digital for access to the NHS register; and the Department of Health

and Social Care for logistic support. SR acknowledges helpful discussion with attendees of

meetings of the UK Government Office for Science (GO-Science) Scientific Pandemic

Influenza – Modelling (SPI-M) committee.

9

Tables and Figures

10

Table 1. Unweighted and weighted prevalence of swab-positivity across nine rounds ofREACT-1 and round 10.

11

Table 2. Estimates of national growth rates, doubling times and reproduction numbers forround 9 to round 10, and within round 10.

12

Table 3. Estimates of regional growth rates, doubling times and reproduction numbers forround 9 to round 10.

13

Table 4a. Unweighted and weighted prevalence of swab-positivity for sex age and region forrounds 10. *We present weighted prevalence if the number of positives in a category is 10 ormore.

14

Table 4b. Unweighted and weighted prevalence of swab-positivity for sex age and region forrounds 9. *We present weighted prevalence if the number of positives in a category is 10 ormore.

15

Figure 1. Prevalence of national swab-positivity for England estimated using a P-spline forall nine rounds with central 50% (dark grey) and 95% (light grey) posterior credible intervals.Shown here only for the period of round 8 to round 10. Unweighted observations (black dots)and 95% binomial confidence intervals (vertical lines) are also shown.

16

Figure 2. Weighted prevalence of swab-positivity by region for rounds 9 and 10. Bars show95% confidence intervals. See Table 4.

17

Figure 3. Neighbourhood prevalence of swab-positivity for rounds 9 and 10. Neighbourhoodprevalence calculated from 4,732 (round 9) and 4,016 (round 10) nearest neighbours in thesample (the median number of neighbours within 30km in the study). Neighbourhoodprevalence displayed as an average of a random sample of 15 participants for eachlower-tier local authority (LTLA). Regions: NE = North East, NW = North West, YH =Yorkshire and The Humber, EM = East Midlands, WM = West Midlands, EE = East ofEngland, L = London, SE = South East, SW = South West.

18

Figure 4. Weighted prevalence of swab-positivity by age groups for rounds 9 and 10. Barsshow 95% confidence intervals. See Table 4.

19

Figure 5. Unweighted proportion of people who had travelled abroad in the previous twoweeks in England and its nine constituent regions for rounds 5 to 10. Vertical line shows95% binomial confidence intervals.

20

Figure 6. Daily hospital admissions in England (solid red line, right hand y-axis) shifted by alag parameter along the x-axis (see below), and daily swab-positivity for all 9 rounds of thestudy (dots with 95% confidence intervals, left hand y-axis) and the P-spline estimate forswab positivity (Solid black line, dark grey shaded area is 50% central credible interval, lightgrey shaded area is 95% central credible interval, left-hand y-axis). Daily hospitaladmissions have been fit to observations from the REACT-1 rounds 1-8 to obtain scaling andlag parameters. These parameter values were estimated using a Bayesian MCMC model:daily_positives(t) ~ Binomial(daily_swab_tests(t), p = daily_admissions(t+lag)*scale). Thediscrete time lag parameter was estimated at 18 (18, 18) days. Note the P-spline is notplotted for the region between round 7 and 8 in which there was a peak in swab-positivitywhile REACT-1 was not active.

21

Figure 7. Daily deaths in England (solid red line, right hand y-axis) shifted by a lagparameter along the x-axis (see below), and daily swab positivity for all 9 rounds of the study(dots with 95% confidence intervals, left hand y-axis) and the P-spline estimate for swabpositivity (Solid black line, dark grey shaded area is 50% central credible interval, light greyshaded area is 95% central credible interval, left-hand y-axis). Daily deaths have been fit toobservations from the REACT-1 rounds 1-8 to obtain scaling and lag parameters. Theseparameter values were estimated using a Bayesian MCMC model: daily_positives(t) ~Binomial(daily_swab_tests(t), p = daily_admissions(t+lag)*scale). The discrete time lagparameter was estimated at 27 (27, 27) days. Note the P-spline is not plotted for the regionbetween round 7 and 8 in which there was a peak in swab-positivity while REACT-1 was notactive.

22